diff options
| author | Gerd Hoffmann <kraxel@redhat.com> | 2023-05-16 11:47:58 +0200 |
|---|---|---|
| committer | mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> | 2023-05-29 11:19:51 +0000 |
| commit | 27727338b2c0e3f50eb0176a1044e903fcb3c3b1 (patch) | |
| tree | a3d3ec392a4738844d7bcfa54d59485ab68fb102 /OvmfPkg/VirtioNetDxe/SnpReceive.c | |
| parent | 9181782c305d08bed369410b45f3d2cae9c0a123 (diff) | |
| download | edk2-27727338b2c0e3f50eb0176a1044e903fcb3c3b1.tar.gz edk2-27727338b2c0e3f50eb0176a1044e903fcb3c3b1.tar.bz2 edk2-27727338b2c0e3f50eb0176a1044e903fcb3c3b1.zip | |
OvmfPkg/PciHotPlugInitDxe: Do not reserve IO ports by default.
Flip the default for IO address space reservations for PCI(e) bridges
and root ports with hotplug support from TRUE to FALSE.
PCI(e) bridges will still get IO address space assigned in case:
(a) Downstream devices actually need IO address space, or
(b) Explicit configuration, using "qemu -device
pcie-root-port,io-reserve=<size>".
In case IO address space is exhausted edk2 will stop assigning resources
to PCI(e) bridges. This is not limited to IO resources, the affected
bridges will not get any memory resources assigned either.
This patch solves this issue by not handing out the scarce IO address
space, which is not needed in most cases anyway. Result is a more
consistent PCI configuration in virtual machine configurations with many
PCie root ports.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Diffstat (limited to 'OvmfPkg/VirtioNetDxe/SnpReceive.c')
0 files changed, 0 insertions, 0 deletions